Exemples de simulations intensives

source('../Fonctions/Simulations.R')
library(ggplot2)
library(dplyr)
library(patchwork)

Scénario sparses

Scénario imaginés dans le cadre du stage pour données sparses, étendus au contexte de données denses.

Graphiques complets

# 10 mesures
data.simu1 <- generation.data(250, 10, 20, "sparse", 1)
data.simu2 <- generation.data(250, 10, 20, "sparse", 2)
data.simu3 <- generation.data(250, 10, 20, "sparse", 3)
data.simu4 <- generation.data(250, 10, 20, "sparse", 4)
graph1 <- plot.data(data.simu1, title = "Scenario 1", legend = F, x = " ", y = "Marqueur", title.pos = 0.5)
graph2 <- plot.data(data.simu2, title = "Scenario 2", legend = F, x = " ", y = " ", title.pos = 0.5)
graph3 <- plot.data(data.simu3, title = "Scenario 3", legend = F, x = "Temps", y = "Marqueur", title.pos = 0.5)
graph4 <- plot.data(data.simu4, title = "Scenario 4", legend = T, legend.name = "Type d'évolution", x = "Temps", y = " ", title.pos = 0.5)
graph.complet_10 <- (graph1 + graph2) / (graph3 + graph4) + 
  plot_layout(guides = 'collect') & theme(legend.position = 'bottom')
graph.complet_10

# 50 mesures
data.simu1 <- generation.data(250, 50, 20, "sparse", 1)
data.simu2 <- generation.data(250, 50, 20, "sparse", 2)
data.simu3 <- generation.data(250, 50, 20, "sparse", 3)
data.simu4 <- generation.data(250, 50, 20, "sparse", 4)
graph1 <- plot.data(data.simu1, title = "Scenario 1", legend = F, x = " ", y = "Marqueur", title.pos = 0.5)
graph2 <- plot.data(data.simu2, title = "Scenario 2", legend = F, x = " ", y = " ", title.pos = 0.5)
graph3 <- plot.data(data.simu3, title = "Scenario 3", legend = F, x = "Temps", y = "Marqueur", title.pos = 0.5)
graph4 <- plot.data(data.simu4, title = "Scenario 4", legend = T, legend.name = "Type d'évolution", x = "Temps", y = " ", title.pos = 0.5)
graph.complet_50 <- (graph1 + graph2) / (graph3 + graph4) + 
  plot_layout(guides = 'collect') & theme(legend.position = 'bottom')
graph.complet_50

# 100 mesures
data.simu1 <- generation.data(250, 100, 20, "sparse", 1)
data.simu2 <- generation.data(250, 100, 20, "sparse", 2)
data.simu3 <- generation.data(250, 100, 20, "sparse", 3)
data.simu4 <- generation.data(250, 100, 20, "sparse", 4)
graph1 <- plot.data(data.simu1, title = "Scenario 1", legend = F, x = " ", y = "Marqueur", title.pos = 0.5)
graph2 <- plot.data(data.simu2, title = "Scenario 2", legend = F, x = " ", y = " ", title.pos = 0.5)
graph3 <- plot.data(data.simu3, title = "Scenario 3", legend = F, x = "Temps", y = "Marqueur", title.pos = 0.5)
graph4 <- plot.data(data.simu4, title = "Scenario 4", legend = T, legend.name = "Type d'évolution", x = "Temps", y = " ", title.pos = 0.5)
graph.complet_100 <- (graph1 + graph2) / (graph3 + graph4) + 
  plot_layout(guides = 'collect') & theme(legend.position = 'bottom')
graph.complet_100

# 250 mesures
data.simu1 <- generation.data(250, 250, 20, "sparse", 1)
data.simu2 <- generation.data(250, 250, 20, "sparse", 2)
data.simu3 <- generation.data(250, 250, 20, "sparse", 3)
data.simu4 <- generation.data(250, 250, 20, "sparse", 4)
graph1 <- plot.data(data.simu1, title = "Scenario 1", legend = F, x = " ", y = "Marqueur", title.pos = 0.5)
graph2 <- plot.data(data.simu2, title = "Scenario 2", legend = F, x = " ", y = " ", title.pos = 0.5)
graph3 <- plot.data(data.simu3, title = "Scenario 3", legend = F, x = "Temps", y = "Marqueur", title.pos = 0.5)
graph4 <- plot.data(data.simu4, title = "Scenario 4", legend = T, legend.name = "Type d'évolution", x = "Temps", y = " ", title.pos = 0.5)
graph.complet_250 <- (graph1 + graph2) / (graph3 + graph4) + 
  plot_layout(guides = 'collect') & theme(legend.position = 'bottom')
graph.complet_250

Graphiques avec NA

SCÉNARIO À RECALIBRER POUR S’ADAPTER À L’AUGMENTATION DU NOMBRE DE MESURES

# Boucle NA Manuelle
source("Simulations.R")
k <- 500
scen <- 1
tx.do <- 0.3
pattern <- "threshMNAR"
do.mean <- numeric(k)
for (i in 1:k){
  data.simu <- generation.data(250, 100, 20, "sparse", scen)
  data.long <- data.frame(id = rep(1:ncol(data.simu$t),
                                   each = nrow(data.simu$t)),
                          temps = as.vector(data.simu$t),
                          y = as.vector(data.simu$y),
                          groupe = rep(data.simu$groupe,
                                       each = nrow(data.simu$t)))
  data.dropout <- dropout(data = data.simu, miss.pattern = pattern, txdo = tx.do, scenario = scen)
  do.mean[i] <- data.dropout$txevent
}
mean(do.mean)

[1] 0.809888

# Graphiques
scen <- 2
tx.do <- 0.3
data.simu <- generation.data(250, 100, 20, "sparse", scen)
data.dpo1 <- dropout(data.simu, "threshMAR", tx.do, scen)
data.dpo2 <- dropout(data.simu, "threshMNAR", tx.do, scen)
data.dpo3 <- dropout(data.simu, "incrMAR", tx.do, scen)
data.dpo4 <- dropout(data.simu, "incrMNAR", tx.do, scen)
graph1 <- plot.data(data.dpo1, title = "Threshold MAR", legend = F, x = " ", y = "Marqueur", title.pos = 0.5)
graph2 <- plot.data(data.dpo2, title = "Threshold MNAR", legend = F, x = " ", y = " ", title.pos = 0.5)
graph3 <- plot.data(data.dpo3, title = "Increasing MAR", legend = F, x = "Temps", y = "Marqueur", title.pos = 0.5)
graph4 <- plot.data(data.dpo4, title = "Increasing MNAR", legend = T, x = "Temps", y = " ", title.pos = 0.5)
graph.na <- (graph1 + graph2) / (graph3 + graph4) + 
  plot_layout(guides = 'collect') & theme(legend.position = 'bottom')
graph.na

Warning: Removed 8641 rows containing missing values or values outside the scale range
(`geom_line()`).

Warning: Removed 8641 rows containing missing values or values outside the scale range
(`geom_point()`).

Warning: Removed 7767 rows containing missing values or values outside the scale range
(`geom_line()`).

Warning: Removed 7767 rows containing missing values or values outside the scale range
(`geom_point()`).

Warning: Removed 16869 rows containing missing values or values outside the scale range
(`geom_line()`).

Warning: Removed 16869 rows containing missing values or values outside the scale range
(`geom_point()`).

Warning: Removed 15365 rows containing missing values or values outside the scale range
(`geom_line()`).

Warning: Removed 15365 rows containing missing values or values outside the scale range
(`geom_point()`).

Scénarios denses

Scénario imaginés pour un contexte de données denses.

data.simu1 <- generation.data(250, 250, 20, "dense", 1, ngroups = 3)
data.simu2 <- generation.data(250, 250, 20, "dense", 2, ngroups = 3)
data.simu3 <- generation.data(250, 250, 20, "dense", 3, ngroups = 3)
data.simu4 <- generation.data(250, 250, 20, "dense", 4, ngroups = 3)
graph1 <- plot.data(data.simu1, title = "Scenario 1", legend = F, x = " ", y = "Marqueur", title.pos = 0.5)
graph2 <- plot.data(data.simu2, title = "Scenario 2", legend = F, x = " ", y = " ", title.pos = 0.5)
graph3 <- plot.data(data.simu3, title = "Scenario 3", legend = F, x = "Temps", y = "Marqueur", title.pos = 0.5)
graph4 <- plot.data(data.simu4, title = "Scenario 4", legend = T, legend.name = "Type d'évolution", x = "Temps", y = " ", title.pos = 0.5)
graph.complet <- (graph1 + graph2) / (graph3 + graph4) + 
  plot_layout(guides = 'collect') & theme(legend.position = 'bottom')
graph.complet